Re: [PATCH 0/3] scripts/spelling.txt: add some spelling pairs and reorder

[Andy Shevchenko]

On Fri, Jun 11, 2021 at 05:48:14PM +0800, Leizhen (ThunderTown) wrote:
> On 2021/6/11 16:10, Andy Shevchenko wrote:
> > On Fri, Jun 11, 2021 at 11:02 AM Andy Shevchenko
> > <andy.shevchenko@xxxxxxxxx> wrote:
> >> On Fri, Jun 11, 2021 at 10:19 AM Zhen Lei <thunder.leizhen@xxxxxxxxxx> wrote:
> >>>
> >>> Add spelling_sanitizer.sh and use it to reorder, then add some spelling
> >>> "mistake||correction" pairs.
> >>
> >> The sorting idea is good, but the order is not.
> >> What you really need is to use language corpus [1] instead. So in such
> >> case you will eliminate false positives (to some extent).
> > 
> > Perhaps I need to elaborate what I meant. The (important) feature of
> > the corpus is sorting by frequency of the word usage. That's what
> 
> That's unlikely. Even if that were the case, no one could figure 'the frequency' out.

What do you mean? It's a natural sorting for the language corpus. First you get
the most used words down to less used.

> > would be the best. Unfortunately I don't know if codespell uses linear
> > search or hash based (i.o.w. does it convert the input file to the
> > Python list() or set() object?).
> 
> I think "spelling.txt" will be reprocessed by the tool.

Of course, and that's what I meant in the parentheses, i.e. to check what data
structure is used behind the scene.

> For the same set of data, if the performance differs significantly or even does not work
> due to the user's input order, the implementation of the tool itself must be problematic.
> 
> So our ranking of "spelling.txt" is how to make it easy to see.
> 
> I don't know if it's right. Pure logical reasoning.

Depends on the above it might increase the ratio of false positives.
Also it depends to the data base (spelling.txt).

What you need here is to make sure that:
 1) tool using arbitrary ordering already (by implying hash table, i.e. set()
    in Python);
 2) the current ordering is arbitrary and doesn't imply any side effects.

With this research conducted and explained, I'm fine with reordering.

> >> [1]: https://en.wikipedia.org/wiki/Corpus_of_Contemporary_American_English

-- 
With Best Regards,
Andy Shevchenko